Training Models PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   Which Google Cloud offering is designed for training tabular models with minimal code, using a workflow-based approach that can be integrated into pipelines?

   • ACloud Composer
   • BCloud Dataflow
   • CTabular Workflows
   • DAgent Platform Custom Training
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A data science team uses Vertex AI custom training to run a distributed TensorFlow job. The job fails after a few minutes with an error indicating that workers cannot communicate with the chief worker. The team verified that the training code is correct and that the data is accessible. What is the most likely cause and solution?

   • AThe cluster is not configured for distributed training; set the cluster spec and TF_CONFIG environment variable correctly.
   • BThe training container does not have internet access; enable public internet access for the job.
   • CThe worker pool's machine type is too small; increase the machine type for all worker pools.
   • DThe workers are running in different regions; reconfigure the job to run all workers in the same region.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   A machine learning engineer wants to run a custom training job using a Docker container with a custom training script on Google Cloud. Which service is designed for this purpose?

   • ABigQuery ML
   • BCloud Functions
   • CAgent Platform AutoML
   • DAgent Platform Custom Training
   Reveal Solution Discussion Select an answer first
5. 9expert · medium

   A team is building a training pipeline that ingests data from multiple sources: a Cloud SQL database, a Kafka topic, and files in Cloud Storage. They want to use Vertex AI custom training. What is the best way to ingest this data into the training job?

   • AExport all data to Cloud Storage in a common format (e.g., TFRecord) before training.
   • BLoad all data into BigQuery and use the BigQuery Storage API.
   • CUse Dataflow to stream data directly into the training job.
   • DHave the training job connect directly to Cloud SQL and Kafka during training.
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A training job on Vertex AI fails with an 'Out of memory' error. The job uses a custom container and reads data from Cloud Storage. The team suspects the data loading is inefficient. Which change is most likely to resolve the issue?

   • AUse the tf.data API with map and batch operations to stream data efficiently.
   • BIncrease the machine type's memory for the training job.
   • CReduce the number of training steps.
   • DDownload all data to local disk before training starts.
   Reveal Solution Discussion Select an answer first